Nextcloud Talk high-performance back-end

[doodlemania2]

@m-si someone may find a way to package it, however, I've found it to be much easier to run a dedicated set of servers for NTHPBE. I offer it to clients for a pretty low fee in either US or DE if you are at all interested.

[mudassar]

hi all, to bring up something from long ago, is there any chance this may pop up?

[doodlemania2]

@mudassar still offering the service for users - ping me on DM if interested.

[avatar1024]

I'm just wondering in light of the Jitsi situation (i.e. hard to update), whether it might be worth reviving this thread starting with the package @doodlemania2 made? Or if any other video call app might be simpler to package and maintain on Cloudron (e.g. @Kubernetes made an initial package for MiroTalk but I've got no clue how it compares with Jitsi or NC talk)

[andreasdueren]

@avatar1024 said in Nextcloud Talk high-performance back-end:

I'm just wondering in light of the Jitsi situation (i.e. hard to update), whether it might be worth reviving this thread starting with the package @doodlemania2 made? Or if any other video call app might be simpler to package and maintain on Cloudron (e.g. @Kubernetes made an initial package for MiroTalk but I've got no clue how it compares with Jitsi or NC talk)

For everyone who wants to host it themselves without too much hassle, I have been successfully using this build on a standalone server for some months now: https://github.com/sunweaver/nextcloud-high-performance-backend-setup

[svtx]

@andreasdueren Aaah, fantastic! Thank you for sharing this repo. It was literal child's play building the HA server with that.

[jdaviescoates]

We really need to have this in Cloudron somehow or another, and sooner rather than later, as it's becoming a standard and expected part of Nextcloud.

Case in point: I just installed fresh instance of Nextcloud on one of my Cloudrons and it immediately gave me this error/ warning about the High Performance Backend being missing:

No High-performance backend configured - Running Nextcloud Talk without the High-performance backend only scales for very small calls (max. 2-3 participants). Please set up the High-performance backend to ensure calls with multiple participants work seamlessly. For more details see the documentation .

[jdaviescoates]

@andreasdueren said in Nextcloud Talk high-performance back-end:

For everyone who wants to host it themselves without too much hassle, I have been successfully using this build on a standalone server for some months now: https://github.com/sunweaver/nextcloud-high-performance-backend-setup

Sounds like a great solution for now, thanks!

@andreasdueren can you confirm or otherwise if this package includes the ability to record meetings? Thanks!

[andreasdueren]

@jdaviescoates It currently does not.

[jdaviescoates]

@andreasdueren OK, thanks. Shame, that's a nice to have too.

[andreasdueren]

I wrote a small script so I can monitor the HPB (turn) backend. Because if it fails, it breaks Nextcloud talk. Posted here in case anyone else wants to easily monitor his turns server.

The cron job runs it once a minute. Each run generates REST credentials from the static auth secret, calls turnutils_uclient to allocate on the TURN server, and then sends a push event to Uptime Kuma containing either TURN allocate succeeded or the last error line from turnutils_uclient. So Kuma sees “up” whenever the allocate completes successfully and “down” when it fails.

[LoudLemur]

If you would like to use the excellent Nextcloud Talk to teleconference with larger numbers of people, it is suggested that you run the HPBE.

There is an AIO Nextcloud release which has the HPBE integrated and it might be a nice idea to switch to this fuller version of Nextcloud in the future.

[james]

Hello @loudlemur
I have merged your topic into this one since it is a duplicate.
I also deleted your AI generated output since this degrades the forum search and potentially creates feedback-loop degradation for AIs.
In other words, AI training on AI output is 'House of Habsburg' like behaviour.

If you wish to share output from an AI in the future, like for example ChatGPT, use the share function to share the AI Chat with a URL.
Example: ChatGPT: What is Cloudron?

[LoudLemur]

@james thanks, will do.

Would setting up hastebin or privatebin on the cloudron demo be good for this, or would it be better off of cloudron altogether?

[james]

Hello @LoudLemur
Since the Cloudron demo server is not persistent, it will get lost there.
The Cloudron https://paste.cloudron.io/ could be used, but also has a policy to delete pastes that are too old.

[charlesnw]

Why not use the excellent MiroTalk for this functionality?

[jdaviescoates]

@charlesnw because Nextcloud Talk nicely integrates with the rest of Nextcloud.

But also quite a few reports on here of people having issues with MiroTalk, even if eventually resolved.